[Bug target/112276] New: [14 Regression] wrong code with -O2 -msse4.2

[Bug target/112276] New: [14 Regression] wrong code with -O2 -msse4.2

[zsojka at seznam dot cz]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112276

            Bug ID: 112276
           Summary: [14 Regression] wrong code with -O2 -msse4.2
           Product: gcc
           Version: 14.0
            Status: UNCONFIRMED
          Keywords: wrong-code
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: zsojka at seznam dot cz
  Target Milestone: ---
              Host: x86_64-pc-linux-gnu
            Target: x86_64-pc-linux-gnu

Created attachment 56469
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=56469&action=edit
reduced testcase

Output:
$ x86_64-pc-linux-gnu-gcc -O2 -msse4.2 testcase.c
$ ./a.out 
Aborted

$ x86_64-pc-linux-gnu-gcc -v
Using built-in specs.
COLLECT_GCC=/repo/gcc-trunk/binary-latest-amd64/bin/x86_64-pc-linux-gnu-gcc
COLLECT_LTO_WRAPPER=/repo/gcc-trunk/binary-trunk-r14-4995-20231029071248-gc6929b08558-checking-yes-rtl-df-extra-nobootstrap-amd64/bin/../libexec/gcc/x86_64-pc-linux-gnu/14.0.0/lto-wrapper
Target: x86_64-pc-linux-gnu
Configured with: /repo/gcc-trunk//configure --enable-languages=c,c++
--enable-valgrind-annotations --disable-nls --enable-checking=yes,rtl,df,extra
--disable-bootstrap --with-cloog --with-ppl --with-isl
--build=x86_64-pc-linux-gnu --host=x86_64-pc-linux-gnu
--target=x86_64-pc-linux-gnu --with-ld=/usr/bin/x86_64-pc-linux-gnu-ld
--with-as=/usr/bin/x86_64-pc-linux-gnu-as --disable-libstdcxx-pch
--prefix=/repo/gcc-trunk//binary-trunk-r14-4995-20231029071248-gc6929b08558-checking-yes-rtl-df-extra-nobootstrap-amd64
Thread model: posix
Supported LTO compression algorithms: zlib zstd
gcc version 14.0.0 20231029 (experimental) (GCC)

[Bug target/112276] [14 Regression] wrong code with -O2 -msse4.2

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112276

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |14.0

[Bug target/112276] [14 Regression] wrong code with -O2 -msse4.2

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112276

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2023-10-29
     Ever confirmed|0                           |1
             Status|UNCONFIRMED                 |NEW

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Confirmed.

[Bug target/112276] [14 Regression] wrong code with -O2 -msse4.2 since r14-4964-g7eed861e8ca3f5

[tkoenig at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112276

Thomas Koenig <tkoenig at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |liuhongt at gcc dot gnu.org
            Summary|[14 Regression] wrong code  |[14 Regression] wrong code
                   |with -O2 -msse4.2           |with -O2 -msse4.2 since
                   |                            |r14-4964-g7eed861e8ca3f5

--- Comment #2 from Thomas Koenig <tkoenig at gcc dot gnu.org> ---
Bisected to r14-4964-g7eed861e8ca3f5 .

[Bug target/112276] [14 Regression] wrong code with -O2 -msse4.2 since r14-4964-g7eed861e8ca3f5

[crazylht at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112276

Hongtao.liu <crazylht at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |crazylht at gmail dot com

--- Comment #3 from Hongtao.liu <crazylht at gmail dot com> ---
Mine, I'll take a look.

[Bug target/112276] [14 Regression] wrong code with -O2 -msse4.2 since r14-4964-g7eed861e8ca3f5

[crazylht at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112276

--- Comment #4 from Hongtao.liu <crazylht at gmail dot com> ---
-(define_split
-  [(set (match_operand:V2HI 0 "register_operand")
-        (eq:V2HI
-          (eq:V2HI
-            (us_minus:V2HI
-              (match_operand:V2HI 1 "register_operand")
-              (match_operand:V2HI 2 "register_operand"))
-            (match_operand:V2HI 3 "const0_operand"))
-          (match_operand:V2HI 4 "const0_operand")))]
-  "TARGET_SSE4_1"
-  [(set (match_dup 0)
-        (umin:V2HI (match_dup 1) (match_dup 2)))
-   (set (match_dup 0)
-        (eq:V2HI (match_dup 0) (match_dup 2)))])

the splitter is wrong when op1 == op2.(the original pattern returns 0, after
splitter, it returns 1)
So remove the splitter.

[Bug target/112276] [14 Regression] wrong code with -O2 -msse4.2 since r14-4964-g7eed861e8ca3f5

[zsojka at seznam dot cz]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112276

--- Comment #5 from Zdenek Sojka <zsojka at seznam dot cz> ---
(In reply to Hongtao.liu from comment #4)
> -(define_split
> -  [(set (match_operand:V2HI 0 "register_operand")
> -        (eq:V2HI
> -          (eq:V2HI
> -            (us_minus:V2HI
> -              (match_operand:V2HI 1 "register_operand")
> -              (match_operand:V2HI 2 "register_operand"))
> -            (match_operand:V2HI 3 "const0_operand"))
> -          (match_operand:V2HI 4 "const0_operand")))]
> -  "TARGET_SSE4_1"
> -  [(set (match_dup 0)
> -        (umin:V2HI (match_dup 1) (match_dup 2)))
> -   (set (match_dup 0)
> -        (eq:V2HI (match_dup 0) (match_dup 2)))])
> 
> the splitter is wrong when op1 == op2.(the original pattern returns 0, after
> splitter, it returns 1)
> So remove the splitter.

Thank you for having a look. I have other testcases, possibly related, that are
being reduced now, failing at -O2 only; so -msse4.2 might not be the only
broken target. I will open a new PR if those will still be broken after this is
fixed; alternatively, I can upload the testcases here, once they are reduced.

[Bug target/112276] [14 Regression] wrong code with -O2 -msse4.2 since r14-4964-g7eed861e8ca3f5

[zsojka at seznam dot cz]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112276

--- Comment #6 from Zdenek Sojka <zsojka at seznam dot cz> ---
Created attachment 56477
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=56477&action=edit
testcase failing at -O2

$ x86_64-pc-linux-gnu-gcc -O2 testcase.c
$ ./a.out 
Aborted

[Bug target/112276] [14 Regression] wrong code with -O2 -msse4.2 since r14-4964-g7eed861e8ca3f5

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112276

--- Comment #7 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by hongtao Liu <liuhongt@gcc.gnu.org>:

https://gcc.gnu.org/g:f5d33d0c790a00cb5240a966dd99df9961bf57a0

commit r14-5025-gf5d33d0c790a00cb5240a966dd99df9961bf57a0
Author: liuhongt <hongtao.liu@intel.com>
Date:   Mon Oct 30 14:05:25 2023 +0800

    Fix wrong code due to incorrect define_split

    -(define_split
    -  [(set (match_operand:V2HI 0 "register_operand")
    -        (eq:V2HI
    -          (eq:V2HI
    -            (us_minus:V2HI
    -              (match_operand:V2HI 1 "register_operand")
    -              (match_operand:V2HI 2 "register_operand"))
    -            (match_operand:V2HI 3 "const0_operand"))
    -          (match_operand:V2HI 4 "const0_operand")))]
    -  "TARGET_SSE4_1"
    -  [(set (match_dup 0)
    -        (umin:V2HI (match_dup 1) (match_dup 2)))
    -   (set (match_dup 0)
    -        (eq:V2HI (match_dup 0) (match_dup 2)))])

    the splitter is wrong when op1 == op2.(the original pattern returns 0,
after split, it returns 1)
    So remove the splitter.

    Also extend another define_split to define_insn_and_split to handle
    below pattern

    494(set (reg:V4QI 112)
    495    (unspec:V4QI [
    496            (subreg:V4QI (reg:V2HF 111 [ bf ]) 0)
    497            (subreg:V4QI (reg:V2HF 110 [ af ]) 0)
    498            (subreg:V4QI (eq:V2HI (eq:V2HI (reg:V2HI 105)
    499                        (const_vector:V2HI [
    500                                (const_int 0 [0]) repeated x2
    501                            ]))
    502                    (const_vector:V2HI [
    503                            (const_int 0 [0]) repeated x2
    504                        ])) 0)
    505        ] UNSPEC_BLENDV))

    define_split doesn't work since pass_combine assume it produces at
    most 2 insns after split, but here it produces 3 since we need to move
    const0_rtx (V2HImode) to reg. The move insn can be eliminated later.

    gcc/ChangeLog:

            PR target/112276
            * config/i386/mmx.md (*mmx_pblendvb_v8qi_1): Change
            define_split to define_insn_and_split to handle
            immediate_operand for comparison.
            (*mmx_pblendvb_v8qi_2): Ditto.
            (*mmx_pblendvb_<mode>_1): Ditto.
            (*mmx_pblendvb_v4qi_2): Ditto.
            (<code><mode>3): Remove define_split after it.
            (<code>v8qi3): Ditto.
            (<code><mode>3): Ditto.
            (<ode>v2hi3): Ditto.

    gcc/testsuite/ChangeLog:

            * g++.target/i386/part-vect-vcondhf.C: Adjust testcase.
            * gcc.target/i386/pr112276.c: New test.

[Bug target/112276] [14 Regression] wrong code with -O2 -msse4.2 since r14-4964-g7eed861e8ca3f5

[crazylht at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112276

--- Comment #8 from Hongtao.liu <crazylht at gmail dot com> ---
Fixed.

[Bug target/112276] [14 Regression] wrong code with -O2 -msse4.2 since r14-4964-g7eed861e8ca3f5

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112276

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |FIXED
             Status|NEW                         |RESOLVED

--- Comment #9 from Richard Biener <rguenth at gcc dot gnu.org> ---
Fixed.